Re: [squid-users] Memory usage question

2015-06-22 Thread Amos Jeffries
On 22/06/2015 5:28 p.m., Alex Samad wrote:
 Hi
 
 UFS or AUFS ? guessing aufs
 

UFA, AUFS, diskd - all the same in this regard.

 Any suggestions on the L1 L2  values, defaults ?

Not particularly. Though others may have preferences. AFAIK the FS
directory index issues in ext2 that L1/L2 exist to resolve have not been
a problem in any OS for many years now.

Amos

___
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users


Re: [squid-users] Memory usage question

2015-06-21 Thread Alex Samad
Hi

UFS or AUFS ? guessing aufs

Any suggestions on the L1 L2  values, defaults ?


On 21 June 2015 at 11:57, Amos Jeffries squ...@treenet.co.nz wrote:
 On 20/06/2015 9:08 p.m., Alex Samad wrote:
 Hi

 Are there any gotchas i need to look out for.
 Also I have allocated a 1T lun to the VM. Whats the best way to
 allocate this do I use 1 cache_dir or multiple cache_dir.

 The usual one UFS based dir per physical drive and no RAID. That can be
 tricky with SAN/NAS based disk and VMs.


 I currently have 3, is there a way to migrate the cache objects in the
 3 into 1 or do I just delete them and bear the cost of re downloading
 them

 If you need to merge them you can set cache_dir to read-only for a
 period. That way Squid will use their content until it becomes too far
 out of date, whiel storing new objects into the witable cache_dir.


 It should not be a big problem/cost to drop the cache anyway. The
 bandwidth to rebuild a cache is far smaller than most people expect.
 Much of the content in a large cache is stale objects waiting
 revalidation or replacement, and all HITs are duplicates by definition.
 The cache fill rate is an exponential/polynomial growth curve with the
 bulk of it being a few seconds/minutes worth of traffic.

 Amos

 ___
 squid-users mailing list
 squid-users@lists.squid-cache.org
 http://lists.squid-cache.org/listinfo/squid-users
___
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users


Re: [squid-users] Memory usage question

2015-06-20 Thread Alex Samad
Hi

Are there any gotchas i need to look out for.
Also I have allocated a 1T lun to the VM. Whats the best way to
allocate this do I use 1 cache_dir or multiple cache_dir.

I currently have 3, is there a way to migrate the cache objects in the
3 into 1 or do I just delete them and bear the cost of re downloading
them




On 19 June 2015 at 21:16, Eliezer Croitoru elie...@ngtech.co.il wrote:
 First goes first...
 Upgrade to 3.5 or 3.4 branch.
 Then try to use top or htop to get a snapshot of the virtual memory and
 resident memory that squid uses.

 Eliezer

 On 19/06/2015 13:19, Alex Samad wrote:

 this is on centos 6.6
 still using the redhat build squid !
 rpm -q squid
 squid-3.1.10-29.el6.x86_64



 ___
 squid-users mailing list
 squid-users@lists.squid-cache.org
 http://lists.squid-cache.org/listinfo/squid-users
___
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users


Re: [squid-users] Memory usage question

2015-06-20 Thread Amos Jeffries
On 20/06/2015 9:08 p.m., Alex Samad wrote:
 Hi
 
 Are there any gotchas i need to look out for.
 Also I have allocated a 1T lun to the VM. Whats the best way to
 allocate this do I use 1 cache_dir or multiple cache_dir.

The usual one UFS based dir per physical drive and no RAID. That can be
tricky with SAN/NAS based disk and VMs.

 
 I currently have 3, is there a way to migrate the cache objects in the
 3 into 1 or do I just delete them and bear the cost of re downloading
 them

If you need to merge them you can set cache_dir to read-only for a
period. That way Squid will use their content until it becomes too far
out of date, whiel storing new objects into the witable cache_dir.


It should not be a big problem/cost to drop the cache anyway. The
bandwidth to rebuild a cache is far smaller than most people expect.
Much of the content in a large cache is stale objects waiting
revalidation or replacement, and all HITs are duplicates by definition.
The cache fill rate is an exponential/polynomial growth curve with the
bulk of it being a few seconds/minutes worth of traffic.

Amos

___
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users


Re: [squid-users] Memory usage question

2015-06-19 Thread Marcus Kool

What is the physical memory size ??
You might want to read the faq on memory:
http://wiki.squid-cache.org/SquidFaq/SquidMemory

Marcus


On 06/19/2015 07:19 AM, Alex Samad wrote:

Hi

I recently push my squid VM memory up to 65G
i pushed up squid usage (i thought) to 40G

squid.conf
cache_mem 40960 MB

cache.log
2015/06/18 22:12:33| Max Mem  size: 41943040 KB
2015/06/18 22:12:33| Max Swap size: 177527808 KB



but it doesn't seem like its using it

  free -g
  total   used   free sharedbuffers cached
Mem:62  5 57  0  0  1
-/+ buffers/cache:  2 59
Swap:1  0  1

again from squid.conf
cache_dir aufs /var/spool/squid 2 16 256
cache_dir aufs /var/spool/squid2 58368 32 256
cache_dir aufs /var/spool/squid3 85000 32 256

Is this just a case that I don't have enough disk to backend the memory usage ?


this is on centos 6.6
still using the redhat build squid !
rpm -q squid
squid-3.1.10-29.el6.x86_64


  ps -u squid -o pid,rss,vsz,cmd | grep squid.con
   PID   RSSVSZ CMD
  21605 839260 916880 (squid) -f /etc/squid/squid.conf


I have slowly added disk space here and I am actually planning on just
inserting 1T lun (this is VMware) and migrating it all over to there.


What am I missing, also on the cache_dir statement is there any way of
specifying 1G as 1G not 1024000. ie can i specify the units ?

Thanks
___
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users


___
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users


Re: [squid-users] Memory usage question

2015-06-19 Thread Eliezer Croitoru

First goes first...
Upgrade to 3.5 or 3.4 branch.
Then try to use top or htop to get a snapshot of the virtual memory and 
resident memory that squid uses.


Eliezer

On 19/06/2015 13:19, Alex Samad wrote:

this is on centos 6.6
still using the redhat build squid !
rpm -q squid
squid-3.1.10-29.el6.x86_64



___
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users